基于群体智能"隐并行性"实现多任务优化已取得一系列研究成果，但任务间频繁的垂直信息传递导致种群异质性过度增加，进而产生信息负迁移消极影响，这也是目前多任务优化领域尚未完全解决的难题之一。针对此问题，首先将粒子群算法(PSO)与多种群演化信息共享机制相结合，然后引入标杆管理思想实现多层级信息迁移及智能涌现，最后通过计算种群多样性指数有效控制信息迁移频率，提出多级信息迁移多任务优化PSO算法(multi-level information transfer multi-task PSO，MITMPSO)。仿真实验表明，通过设置合理的信息迁移阈值，MITMPSO能在多项式时间内显著提高多任务高维函数优化、多任务多约束函数优化以及多任务二元离散优化问题的求解质量，加快各优化问题的收敛速度。
Multi-level information transfer multi-task PSO based on population diversity control
A series of research achievements have been made in multi-task optimization(MTO)based on the implicit parallelism of swarm intelligence.However,the frequent vertical information transfer between tasks leads to excessive increase of population heterogeneity,resulting in the negative impact of information migration,which is also one of the problems that has not been completely solved in the field of MTO.Firstly,PSO and multi-population evolution information sharing mechanism are combined,then the idea of benchmarking management is introduced to realize multi-level information migration and intelligent emergence,finally the frequency of information transfer is effectively controlled by calculating the population diversity index,and the multi-level information transfer multi-task PSO(MITMPSO)is proposed.Experimental results show that the MITMPSO can significantly improve the solution quality and accelerate the convergence speed of multiple high-dimensional functions,multiple multi-constraints functions and multiple binary discrete optimization problems concurrently in polynomial time by setting a reasonable information migration threshold.
multi-task optimizationparticle swarm optimization(PSO)multi-level information transferpopulation diversity control
万方数据
维普
